Categorical Data and Generalised Linear Models

Description

This course will explore biostatistical applications of generalised linear models with an emphasis on underlying theoretical issues, and practical interpretation of the results of fitting these models. This course is offered in conjunction with the Biostatistical Collaboration of Australia (BCA).

Availability

Distance Education - Callaghan

  • Semester 2 - 2015

Learning Outcomes

1. Understand the major theoretical aspects of generalised linear models

2. Appreciate regression modelling strategies for generalised linear models, including estimation issues, choice of models, prediction and goodness of fit of a selected model

3. Be proficient in the analysis of binary outcome data, either from a standard study design or from a matched study design

4. Be capable of analysing ordered and unordered categorical outcomes using simple measures of association and complex regression models

5. Be capable of analysing count data whether it satisfies standard distributional assumptions or whether it is over-dispersed.

Content

Generalised linear models are a family of models with a unified theory for estimation and inference, including as special cases many of the commonly-used methods in the analysis of health data. Because of the central importance of binary outcomes in epidemiological studies, a thorough grounding in relevant methods for 2x2 and 2xk tables will be given and this will extend naturally into logistic regression for a binary outcome as a special case of generalised linear modelling. A full introduction into measures of association and modelling techniques for ordinal outcomes will be presented. A grounding in methods for analysing count data will be provided. Techniques for dealing with matched data, e.g from case-control studies, will also be introduced.

Requisites

This course is only available to students enrolled in the Graduate Diploma in Medical Statistics or Master of Medical Statistics programs.

Assumed Knowledge

Epidemiology (EPID6420); Mathematical Background for Biostatistics (BIOS6040); Principles of Statistical Inference (BIOS6050); Probability and Distribution Theory (BIOS6170); Linear Models (BIOS6070) Co-requisite. Please note, Program Coordinator approval is required for taking CDA and LMR simultaneously.

Assessment Items

Written Assignment: Essays / Written Assignments

Written Assignment: Worked Exercise

Online Learning Activity: Online Discussion

Contact Hours

Self-Directed Learning

Self-Directed 8 hour(s) per Week for Full Term

8-12 hours total study time per week